A couple of weeks ago, after reading yet another piece of high-minded marketing copy, full of words like hand-crafted and artisan, a silly verse popped unbidden into my head:
This is not the renaissance.
You are not an artisan.
Go around to the back door,
you’re a smelly tradesman.
So long as we’re pretending that we’re rediscovering an early-modern work ethic, I think I can call myself a bard and allow myself a bit of anachronistic doggerel.
Thinking through the implications of the whole artisan-crafts-guilds meme in the future-of-work debates led me to an odd conclusion: the future is significantly brighter (or less bleak) than people realize. So long as you stop thinking in terms of crafts and aim to practice a trade instead, there is more work for humans than people realize.
What’s the difference? It’s the difference between bards and chimney-sweeps.
Conspicuous Production
The future of work looks bleaker than it needs to for one simple reason: we bring consumption sensibilities to production behavior choices. Even our language reflects this: we “shop around” for careers. We look for prestigious brands to work for. We look for “fulfillment” at work. Sometimes we even accept pay cuts to be associated with famous names. This is work as fashion accessory and conversation fodder.
We can think of this as conspicuous production, by analogy to conspicuous consumption. First-world artisan tendencies take this to a logical extreme.
When you subconsciously think of work as something you consume for pleasure, you end up with a possibly irrational (economically speaking) attraction to artisan work. Even those who don’t actually end up as artisans choose work the way they choose cars, jewelry or handbags, over-valuing things like resume-value and exposure-value.
The result is a misguided analysis of the impact of computers and automation that makes us think the future of work is much darker than it is.
What’s the difference between a tradesman and an artisan? Think chimney-sweep versus bard as the extremes of the spectrum. Both are archetypes that mostly disappeared with late industrialization in the early twentieth century, thanks in part to automation, but there the similarities end.
One fulfilled a critical economic function by engaging in unpleasant and inconspicuous production. The other fulfilled a non-critical economic function in the economy by engaging in pleasurable and conspicuous production .
One generated a higher, less volatile income, but with little potential for upward mobility, the other generated a lower, more volatile income, but with more potential for upward mobility.
The median chimney sweep did better than the median bard, but the best bards did better than the best chimney-sweeps (by finding favor with a king for instance). Since this was before mass media, bard reward distribution was not as skewed as it would become, but it was still skewed.
The emerging future of work does resemble pre-modern patterns of labor organization in a few key ways, but most of us are going to turn into digital-era chimney sweeps rather than bards. And this is a good thing.
The difference between bard work and chimney-sweep work is that it far easier to convince yourself that a relaxing hobby is actually real work. It is a kind of gollumizing effect: behavior that makes you atrophy psychologically.
What makes it worse is that in an economy based on a fiat currency, shareholder value maximization and deficit spending, the capacity to generate an income does not necessarily imply that meaningful work is being done, either in a subjective psychological sense (it helps you evolve rather than atrophy) or economic sense (net wealth is being created rather than consumed or harvested). You might even end up having to pay to do real work.
Since income being generated at an individual level is not a reliable indicator that work being performed, I prefer a different distinction: schlep work versus sexy work. If there is schlepping involved, it is more likely to be real work. If there are sexy elements involved, it is more likely to be conspicuous production pretending to be work.
Schlep Work and Sexy Work
I first broached this topic in my 2010 post, Tinker, Tailor, Solider, Sailor, where I noted that worker archetypes seem to fall into two categories in every era. The dull-dirty-dangerous category and the (potentially) sexy-lucrative-powerful category. Let’s simplify the labels into schlep work and sexy work.
Sexy work, such as being a bard, is work that:
- humans find easy to enjoy
- easily catalyzes mindful absorption while learning (flow)
- is easy to value as a status currency
- is good raw material for social identity formation
People who seek sexy work are often members of what I called the Jeffersonian middle class in an earlier post — motivated by creative self-expression and a sense of personal dignity rather than economic survival.
The first three attributes are self-explanatory. By social identity, I mean the part of your self-perception that is derived from what you think others think of you. Sexy work is attractive to those who like their social identity to be harmoniously integrated within itself (what your mom thinks of you and what your boss thinks of you are not in conflict) and with your private identity (you don’t feel misunderstood). There is consensual external validation of your internal sense of self-worth. You feel authentic.
Sexy work is easy to enjoy, learn, value and integrate into your identity, primarily because it is downhill psychological work: it is the cognitive equivalent of muscular atrophy. You have to choose to make it hard for yourself. You can cash out some status and attention even if you’re not making any money. It does not test your sense of self-worth significantly.
Schlep work has the opposite characteristics along all four vectors. It is harder to enjoy, learn, value and integrate into your identity, primarily because it is uphill psychological work for a social species. It is hard whether or not you want it to be. It is hard to cash out status and attention even if you’re making good money. It tests your sense of self-worth every day.
Somehow, over the past decade, we’ve gone from a useful heuristic (“focus on your strengths” and “find flow”) down a slippery slope of use-with-caution ideas (“work smart, not hard” and “follow your passion”) to the idea of work as a kind of consumption that should be chosen based on the pleasure one can derive from it.
Sexy/schleppy is to my mind, the most natural way to break down human preferences for work. They arise from fundamental desires and aversions. In choosing consumption behaviors or conspicuous production, we tend to feed desires and starve aversions. In schleppy work, we do the opposite: we defer gratification and accept, even seek out, a degree of pain based on the no-pain-no-gain heuristic. A little nudge from a plausible “play to your strengths” philosophy is enough for us to choose the easier way.
Unfortunately, the entire current conversation around work is confused because we prefer a less meaningful distinction, creative vs. uncreative.
Creative versus Sexy
I started rethinking my sexy-versus-schlep ideas in light of the emerging debate around jobs, and especially the question of whether the rise of smart machines might lead to a permanent loss of work for humans.
To address this question correctly, we need to stop deluding ourselves about the word creative.
In reading books like the rather banal Race Against the Machine and articles in the same vein, I was struck by an underlying assumption in much of what is being said: that all “non-creative” work is destined to be taken over by automation. So the anxiety around jobs reduces to anxiety around how creative computers and robots can get, and whether there is enough leftover “high-end creative work” to go around for humans.
Defining “creative” is an interesting challenge, but beside the point.
This is because when you actually poke at what people think of as creative — the broader universe around prototypical categories like fine art, rock music or programming — you realize they don’t really mean creative. They mean sexy. The “creative” attribute (whatever its subtle definition might be) is actually an optional extra. Push comes to shove, that’s an attribute people are pretty willing to give up, so long as the four key attributes are preserved (easy to enjoy, easy to learn, easy to value in a status economy, and easy to integrate into an “authentic” social identity).
In other words, we’re more afraid of machines taking away our social status than our jobs. This might seem like an obvious point. After all, most status-conscious people have strong feelings about what work is “beneath” them, but with machines in the picture, the point gets considerably more subtle.
People substitute creative for sexy in describing their aspirations (to themselves and others) because it sounds less narcissistic. If you seek sexy work, you could be viewed as self-absorbed, entitled and attention/status seeking.
If you pretend it is creative work, you’re suddenly God’s gift to the world, basking in the gratitude, admiration and adoration of all simply for existing.
This is one reason vanity startups, garage bands, indie coffee shops and boutique handbag design businesses proliferate. The valuation of Apple is a good proxy for the valuation of sexy work in our economy, since the company effectively panders to sexy-work seekers (it is revealing that Apple’s stronghold is in consumer markets, while Microsoft’s stronghold is in enterprise markets, but I won’t go down that bunny trail).
So when people talk about saving work or jobs, they mostly talk about saving sexy, income-generating conspicuous production packaged as creative work, in a debt-fueled de facto leisure society. Since few people actively aspire to do the schlep work anyway, we don’t poke much at the consensus view that it can be automated away.
I think this conclusion is premature and in fact mistaken. Just because sexy work is the kind we want to save doesn’t mean it is the kind that is easier to save. In fact it is harder to automate schlep work, which we grievously misunderstand.
To understand what makes work easy or hard to save from computers, robots and automation, we have to consider work from the point of view of machines.
Algorithmically Scalable Work
Machines don’t see the world in sexy-versus-schleppy, creative-versus-uncreative or production-versus-consumption terms. For the moment, they don’t consume at all.
For the moment, computers only produce. The world according to computers (and by extension, robots and soon all machines) offers two kinds of work: algorithmically scalable and algorithmically unscalable.
Algorithmically scalable work typically involves a small variety of typical cases that require a sort of bounded-variety repeatability in the tasks being accomplished. A kind of repetition whose logic can be captured within a very compact process description: an elegant algorithm.
These tractable universes, on which elegant algorithms operate, don’t emerge magically out of messy realities. They emerge because humans work to either censor out or encode a million little exceptions, corner cases and arbitrary domain-specific details. They emerge because humans work around problem regimes that resist generalization and simple automated learning models. Programming is theory building, and computers work not with reality, but with the theories we construct for them to inhabit. They need us to define the base categories through which they see the world.
In algorithmically scalable work, machines need very little help from humans to do a lot. In algorithmically unscalable work, they need a lot of help to do much less. Whether what they do is sexy or creative in human terms is besides the point. What matters is how truly repeatable the defined tasks are.
Almost all our confusion about automation can be traced to a single sloppy conflation: between algorithmically scalable/unscalable and schleppy/sexy. We do this by inappropriately defining the word repetitive as “whatever humans find boring.”
Just because we don’t want to do certain kinds of work, doesn’t mean machines are better at it. They might be worse.
Machine Repetition and Human Boredom
We make the mistake of thinking that just because computers do bounded-variety, repetitive information work very well, that they can do anything that seems repetitive (boring) to humans very well.
But when we’re talking complex systems-level schlepping, like the refining of crude data from disparate information systems, there are rarely any elegant algorithms. Just dozens or hundreds of arbitrary details, small fixes, one-time operations, error corrections and so forth. Humans think of it as repetitive work, but it isn’t. It is hundreds of similar, but not identical, special cases that are easy (if tedious) for humans to handle, but resist general attacks via elegant algorithms.
In fact I suspect the amount of messy and non-repetitive but critical detail determines the amount of human work a domain can sustain.
The human share of the work pie isn’t the gap between machine creativity and human creativity. The real human share of the work pie is the gap between machine repeatability and human boredom.
Human work in the digital age is not about the sideshow of faux-creative artisan urges and conspicuous production. It is about accommodating messy variety that requires non-algorithmically-scalable work. Work that is not worth automating. This is a combination of work being hard to automate and low returns to the automation due to limited algorithmic scaling potential (a good example is tax software for parts of the tax code that change very frequently).
Whether or not it is sexy is irrelevant, since machines don’t compete for status or strive to form “authentic” social identities. Whether or not it is creative is also irrelevant since we are usually being insincere when we talk about creativity.
The big point here is that computers are an industrial age technology. Algorithms are codified processes that deliver economies of scale in narrowly circumscribed kinds of information work. They are information processing assembly lines.
There is deep uniformity to the output of algorithms just as there is deep uniformity to the output of assembly lines in China producing coffee mugs. The Big Secret of the Big Data revolution, which is often characterized as teaching computers to handle “volume, velocity and variety,” is that the last of the triad involves far more unsexy human schlepping than the first two.
We forget this obvious point that computers are an economies-of-scale technology because the processes are very high-resolution. So we mistake fine-grained combinatorial variety (think Starbucks drinks or the dizzying variety of cellphone plans) for the sort of true variety that computers still leave humans to handle. It’s another matter that we don’t like having this unsexy work left for us.
Speaking of true variety and coffee mugs, let’s get back to artisans. A class that includes people who think producing hand-crafted coffee mugs in Portland at a cost of $20 apiece is work rather than conspicuous production, and morally superior to $0.50 mugs mass produced in China.
Isn’t there true variety there?
No. The key is an idea called requisite variety that sexy-work seekers take great pains to avoid thinking about.
Requisite Variety and the Artisan Delusion
Aspiring artisans seek sexy work at small-and-local scales. They reject mass celebrity and status in a global culture, but still crave local celebrity and status (they call it “being respected in the community”). They still look to engage in conspicuous production. They are as prone to deluding themselves that sexy is creative as wannabe actors.
How do they do this?
They do this by confusing economically essential variety (such as handling all the potential variety and ongoing evolution in an online payment system) with economically optional variety (such as uniqueness in hand-crafted coffee mugs). This is the artisan delusion.
If the uniqueness in the product mainly makes the producer feel more special and unique, without leading to profitable differentiation, it’s the optional kind, like latte art.
The former is variety that must be handled to make a market profitable. Essential variety exists in even the most low-end, mass-produced version of an economic good. Optional variety only matters, if it matters at all, in premium niches that can only sustain a few producers. When too many producers swarm into these niches, a lottery economy is created and customers essentially enjoy free variety sustained by a churn of deluded producers offering under-priced goods.
Often, the optional variety doesn’t matter even to the most refined customers (who care least about what they get in return for marginal dollars). So an unprofitable amount of marketing effort must be expended simply to convince people to care about distinctions that make no difference.
Most of us would reject the conclusions of a data science project that dropped key data sources or a payment system that failed to handle common methods of payment. Most of us are completely happy drinking reasonably good, non-pour-over coffee out of unremarkable mass-produced mugs from China. Unless of course, we can get the artisan stuff at the same price.
Economically essential variety is related to what systems theorists call requisite variety; irreducible domain variety that must be handled to achieve baseline performance. Artisanal variety is often the opposite of requisite variety: noise. Information that is, at best, an element of consumption pleasure (much of it accruing to the producer rather than the nominal consumer) and somewhere between inconsequential to actively distracting in profitable production. Variety that we are predisposed to ignore as consumers or actively eliminate as producers.
The artisan delusion is important because almost everything artisans want to do — all the local-and-sexy work — is actually algorithmically scalable once you filter out the noise. There just isn’t much requisite variety there. Which means it is more vulnerable to being taken over by post-industrial modes of automated production, not less. Because software makes assembly lines more capable, not less.
This is a harsh truth to accept. So much so that artisans and their small-and-local advocates often go to great lengths to manufacture environmental and other non-economic justifications in the hope of creating structural distortions that neutralize their sexy-work biases. While there are certainly legitimate concerns of this sort around every sort of scaled production (whether industrial or post-industrial), there is often a lot of bad-faith overstatement of such concerns.
An example of this is the case of a disposable artisanal container traditionally used in India to serve tea, kulfi ice-cream and a variety of other street foods. The container, called a kulhar, is an unglazed earthen pot usually made in small kilns in the small-medium scale sector and smashed into shards after use.
These pots are massively damaging to the economy and environment, because they require fertile topsoil to make (so desperate farmers during drought years are often good sources for raw material rights), and are effectively a kind of crude glass that is the opposite of recyclable. They are arguably worse than paper or plastic in terms of landfill degradation time.
Yet, because kulhars look more artisanal than paper or plastic cups (and have historically been associated with small-and-local vendors rather than modern industry), for a while the Indian middle class was championing them as being “better for the environment.”
A Towering Babel of Confusions
So let’s take stock: we’ve schlepped though a series of subtle but pervasive confusions and delusions that cloud the future of work debates and the impact of computers.
- The application of consumption sensibilities to production activity choices
- The assumption that income generation is the same as work in a psychological sense
- The sexy vs. schleppy distinction that emerges as a result
- The creative versus uncreative red herring that we use to disguise narcissistic sexy-work seeking
- The resultant conflation of creative with what humans are supposedly “still better at”
- The obscuring of the algorithmically scalable/unscalable distinction that actually matters
- The conflation of repetitive and boring resulting in a mis-characterization of schlepping
- The failure to recognize that algorithms are an economies-of-scale technology
- The failure to recognize the distinction between true and combinatorial variety
- The failure to distinguish between requisite and optional variety
- The rise of unprofitable boutique sectors that rely on “educating” consumers to care about distinctions that don’t make a difference
- The resulting unwinnable “race against the machine” to retain “creative” human work at the edge of this optional-variety economy
- The overall result: under-resourced schlep-work sectors that need humans and overcrowded sexy-work sectors
This is a formidable multi-dimensional space. We’ve got sexy/schleppy, algorithmically scalable/unscalable, creative/uncreative, high/low requisite variety, profitable/lossy marketing.
But this whole towering babel of confusions falls apart if you discard anthropocentric categories (creative versus uncreative) and human desires (sexy versus schleppy) and distortions resulting from the nature of fiat economies (income versus work) and simply break down work according to what computers qua computers are good at in terms of their own intrinsic nature as algorithmic information factories delivering economies of scale.
If you actually look at the work computers leave for us — supporting algorithmically unscalable information work — you will see that it is a far larger category than the “sexy that can be packaged as creative” subset that we are racing desperately to save. It may still not be enough to keep everybody productively employed, but there is certainly more to do than we think there is.
The easiest way to appreciate the emerging human condition to adopt a couple of new metaphors for machines: machines as children and humans as intestinal flora.
Machines as Children, Humans as Intestinal Flora
First, machines are like children. The opposite of the overlord personification we’ve been encouraged to adopt by science fiction.
Like parents, we have to let them have the fun while we child-proof the environment (sanitize their inputs) and clean up after them (do whatever they are too clumsy to do and clean up any messes they create). They may not (yet) crave status or social identities, but they certainly look for easy-to-learn high-flow tasks: algorithmically scalable work (which is a sort of aspie-sexy choice of work if you think about it).
Second, humans are like the intestinal flora in the body of technology. I don’t recall where I first heard this analogy, but it isn’t original to me.
The data refining example illustrates the second analogy: without intestinal fauna, humans and other animals could not digest many nutrients. Even dish-washing is hard for machines, so we have to pre-clean to some extent before loading dishwashers.
Without humans inhabiting their guts, technological systems cannot process much of the arbitrariness of the world (Amazon’s Mechanical Turk illustrates this most dramatically at scale). This is requisite-variety schlepping, for which we can expect to be paid in proportion to the true creativity involved (which may not be easy to convert into sexy status currencies).
Put the two together and you get a view of technology as a giant child we’ve given birth to, that is probably never going to grow up and take care of us in our retirement. Instead, we’ll have to live in its guts and take care of it forever, doing the complex schlep work it cannot do.
So one way to understand the historical change transforming human society today is by analogy to parenthood. Life as we know it is over. It’s all about taking care of the big, messy brat now (this is another way of describing what I called hackstability in an earlier post). It likes to play chess, drive cars, fly fighter planes, make coffee mugs, win at Jeopardy, compose music and maybe in the future design clothes, do elegant mathematics and so forth.
So what is it like to do thankless parenting work like data refining, while living in the guts of a giant, developmentally disabled savant child?
It’s chimney-sweep work. Not bard work. Thankless parent work. But work that just might help us evolve rather than atrophy.
The Fate of Artisans
If you want to be a bard, prepare for misery. And not the soul-uplifting kind either.
Artisans are going to have a bad time in the next few decades. They will spin their wheels trying to sell nonessential variety just out of reach of machines, that require unprofitable amounts of customer-education marketing. They will hawk under-priced artisanal coffee, food, clothing, jewelry and handbags to a shrinking class of consumers with enough discretionary income.
This phenomenon isn’t new. Photography — another savant-child machine capability — put a lot of painters out of work by taking away the most reliable and profitable supply of requisite-variety work (what we call photo-realistic art today, which I find deliciously ironic).
In every sexy-work market, automation takes away the most profitable irreducible-variety segment and leaves behind a pure attention economy segment subject to the highly random celebrity and fashion dynamics of the Internet age. This follows from the fact that people choose sexy work mostly due to conspicuous production (status and identity) considerations. They are effectively working for attention, not money.
But much of the attention (which is the scarce commodity all sexy sectors compete over) is cornered by a few at the top, leaving dregs for the rest. The Internet merely creates pocket change in the long tail and more churn in the short head. It doesn’t really change industrial-age winner-take-all dynamics.
Certainly, there is room for a few artisan-of-the-week spots in the hand-made coffee mug sector. Every year, maybe a few hundred such artisans will have a profitable year. The rest of their careers will be spent waiting for the next break. The majority will have no breaks at all and crash out of the sector, to be replaced by new hopefuls.
The role of technology in sexy work is to take away the algorithmically scalable, high-requisite-variety market segments that actually generate profits, leaving behind a casino economy for a class that is destitute in the median case. It is democratization in the sense of turning an unfair lottery that you can at least game with some cleverness (such as pitching a gullible movie producer with more money than taste) into a true lottery that you cannot.
The Fate of Tradesmen
The chimney sweep represents schlep work. Dull, check. Dirty, check. Dangerous, check. No bard options in the posterity memeplex economy? Check.
Today, the dangerous part is being engineered out of sector after sector, starting with drones replacing human pilots in war zones. We value human life in ways that cannot be easily modeled in economic terms, so we allocate more wealth to eliminating dangerous than a free market would (whether that allocation is effective or not is a different matter). But dull and dirty enjoy no such moral protection from economic logic, so long as they can be rendered safe.
Fortunately, machines are not quite as good at carving away the profitable parts of schlep work as they are with sexy work. Which means that when you subtract the algorithmically scalable varieties of dull and dirty work, you’ll likely have more than a casino economy left over for humans.
The early example of information-age chimney sweep work are just emerging: data cleaning, image interpretation, human customer service as a differentiator from voice-prompt hell, various kinds of machine repair. Some of these categories will go away, but I suspect new kinds of schlep work will emerge faster than old kinds vanish. That’s what happened the last time around.
Chimney sweeps emerged with early industrialization because more people had coal-heated homes with narrow chimneys in every room. Those in the trade began life as small children climbing down chimneys until they were too big. If they didn’t die of cancer, they had a shot at becoming master sweeps with their own troops of children.
Regulation, appropriate machinery and the slow replacement of coal with gas and electricity eliminated the profession.
But before the trade declined, it helped establish the industrial age middle class, marked by varied kinds of schlep work.
We are just beginning to discover the schlep work in the information economy. From solar panel installers to driverless car debuggers, several schleppy professions are starting to emerge. Those who are fixated on saving sexy work are most likely to miss schleppy opportunities.
What unites all these trades is that they accept roles based on kinds of schlepping that machines are bad at rather than insisting on work that humans like to do.
I think of them as forming an emerging Hamiltonian middle class — a class that accepts and adapts to large-scale technological systems as a part of life (the kind that Alexander Hamilton promoted in early America). Unlike the Jeffersonian middle class, the Hamiltonian middle class is willing and able to redefine its identity and evolve with machines rather than remaining attached to a static, romanticized notion of what it means to be human.
I’ll talk about the two middle classes another day. I am still working this stuff out. Including figuring out where blogging fits in the grand scheme of things. It’s less sexy and more schleppy than it looks, so there’s hope for me yet.
Thanks to Nick Pinkston and Kartik Agaram for helpful conversations.